{"paper_id":"ecd2f273-422c-49a1-aa77-f9749669c932","body_text":"1 \nNo evidence that visual impulses enhance the readout of retrieved long-term 1 \nmemory contents from EEG activity 2 \nSander van Bree1,2,3,4*, Abbie Sarah Mackenzie1, Maria Wimber1,2 3 \n1Centre for Cognitive Neuroimaging, School of Psychology and Neuroscience, University of Glasgow, Glasgow, 4 \nUnited Kingdom 5 \n2Centre for Human Brain Health, School of Psychology, Birmingham, United Kingdom 6 \n3Department of Medicine, Justus Liebig University, Giessen, Germany 7 \n4Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany 8 \n*Correspondence: sandervanbree@gmail.com 9 \n 10 \n 11 \nAbstract 12 \nThe application of multivariate pattern analysis (MVPA) to electroencephalography (EEG) data allows 13 \nneuroscientists to track neural representations at temporally fine-grained scales. This approach has 14 \nbeen leveraged to study the locus and evolution of long -term memory contents in the brain, but a 15 \nlimiting factor is that decoding performance remains low. A key reason for this is that processes like 16 \nencoding and retrieval are intrinsically dynamic across trials and participants, and this runs in tension 17 \nwith MVPA and other techniques that rely on consistent ly unfolding  neural code s to generate 18 \npredictions about memory contents. The presentation of visually perturbing stimuli may experimentally 19 \nregularize brain dynamics, making neural codes more stable across measurements to enhance 20 \nrepresentational readouts.  Such enhancements, which have  repeatedly been demonstrated in 21 \nworking memory contexts, remain to our knowledge unexplored in long -term memory tasks. In this 22 \nstudy, we  evaluated whether visual perturbation s—or pings—improve our ability to predict the 23 \ncategory of retrieved images  from EEG activity during cued recall. Overall, our findings suggest that 24 \nwhile pings evoked a prominent neural response , they did not  reliably produce improvements in 25 \nMVPA-based classification across several analyses. We discuss possibilities that could explain these 26 \nresults, including the role of experimental and analy sis parameter choices  and mechanistic 27 \ndifferences between working and long-term memory. 28 \nKey words : Long -term memory, MVPA, decoding, EEG, ping, visual impulse , perturbation, brain 29 \ndynamics  30 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 2 \nIntroduction 31 \nA central question in memory research  is how the brain retrieve s information stored in long -term 32 \nmemory (LTM) in the service of adaptive behaviour . This research topic has inspired work from  a 33 \nvariety of angles , involving  different experimental protocols and methods—including neuroimaging 34 \nmodalities. Electroencephalography (EEG) and magnetoencephalography (MEG ) have proven  an 35 \nintegral part of this project  because they capture brain dynamics on a sub -second resolution. Such 36 \ngranularity is crucial, given that memory retrieval typically unfolds on the order of seconds, with  the 37 \nneural cascades underpinning memory retrieval evolving even faster (Staresina & Wimber, 2019). 38 \n To study the evolution of retrieved contents in the brain, one widely pursued family of 39 \ntechniques is multivariate pattern analysis (MVPA)—more broadly known as classification or decoding 40 \n(Haxby et al., 2014; Grootswagers et al., 2017) . These tools extract and upweight signal dimensions 41 \nthat robustly covary with retrieved memory contents, effectively boosting the signal-to-noise ratio of  42 \nassociated neural activity. MVPA has been successfully used to enrich our understanding of memory, 43 \nincluding how information is encoded (Fritch et al., 2020; Kragel et al., 2017; Kuhl et al., 2012) , 44 \nconsolidated (Deuker et al., 2013; Maguire, 2014; Schreiner et al., 2021) , and reinstated during 45 \nmemory recall (i.e., pattern completion; Danker & Anderson, 2010; Favila et al., 2020; Rissman & 46 \nWagner, 2012; Xue, 2018). 47 \n Despite such advancements, the decoding of long-term memory contents in electrophysiology 48 \ndata typically remains only slightly above chance, impairing our ability to study the evolution of neural 49 \npatterns of interest. One reason for this limitation is that memory processes and their associated brain 50 \nactivity are highly dynamic, which results in variable patterns across trials and participants (ter Wal et 51 \nal., 2021; Madore & Wagner, 2022). Indeed, MVPA and most other EEG-based analyses rely for their 52 \nrobust predictions on the existence of a detectably constant cascade of neural patterns across 53 \nmeasurements (van Bree et al., 2022) . This clash between variability in neural processes on the one 54 \nhand and the constancy assumption of our analyses on the other may cause us to miss 55 \nrepresentations of interest, or to obtain different results depending on what experimental event we 56 \ntimelock EEG data to  (e.g., retrieval cues vs button presses; Linde-Domingo et al., 2019 ). A factor 57 \nthat further hampers our ability to robustly decode representations is that retrieval comes with fainter 58 \nneural patterns to begin with compared to perception (Favila et al., 2020; Pearson et al., 2015; Favila 59 \net al., 2022) . Together, these points invite creative techniques that improve our ability to infer long-60 \nterm memory representations from dynamic brain activity. 61 \nIn this study, we explore a perturbational method that has the potential to mitigate two issues 62 \nat the same time: low signal fidelity at the level of measurement, and variability in neural processing 63 \ndynamics. Specifically, in this EEG study we evaluated whether the presentation of a high contrast 64 \nvisual stimulus—henceforth referred to as a “ping” —during LTM retrieval enhances the readout of 65 \nsignatures of retrieved content. In motivating the hypothesis that pings boost the decodability of LTM 66 \nrepresentations, we buil t directly onto recent successful efforts in the domain of working memory  67 \n(WM). In that context, pings have been used to enhance the decodability of the orientation (Wolff et 68 \nal., 2015, 2017, 2020; Ten Oever et al., 2020; Yang et al., 2023) and colour (Kandemir et al., 2023) of 69 \nobjects actively maintained in WM, as well as anticipated target locations (Duncan et al., 2023) . A 70 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 3 \npreliminary explanation for these findings is that pings induce a robust evoked response that interacts 71 \nand indeed boosts the footprint of active neural representations, enhancing their SNR (Barbosa et al., 72 \n2021). Specifically, pings may regularize neural dynamics across trials and participants by producing 73 \na phase reset of brain oscillations that coordinate information processing  across neuronal 74 \npopulations. In support of this , visual stimuli presented during memory tasks  have been shown to  75 \nreset the phase of low-frequency brain oscillations  that are implicated in encoding and retrieval  76 \n(Rizzuto et al., 2003; Haque et al., 2015; audiovisual stimuli in Cruzat et al., 2021). Thus, by inducing 77 \npings at experimentally controlled moments, researchers may gain a level of control over variability in 78 \nsynchronized activity across information-coding neurons, making their dynamics more similar across 79 \nmeasurements to improve the predictive power of MVPA. 80 \nImportantly however, while ping-based methods have been shown to work in WM contexts, to 81 \nour knowledge it has not been explored whether they generalize to LTM research in which information 82 \nis retrieved from stored representations.  The purpose of this study  then, is to systematically explore 83 \nthe possibility that pings can enhance the readout of reactivated long -term memory contents. To this 84 \nend, we presented participants with pings as memory processes were actively engaged during cued 85 \nrecall, evaluating whether retrieved representations are more robustly discernible after ping onset. On 86 \nthe whole, we find no compelling evidence that pings boost the classification of retrieved image pairs  87 \nfrom EEG activity. 88 \n 89 \nMethods 90 \nParticipants 91 \nWe recruited thirty-three volunteers (22 women, Mage = 23.8 years, SDage = 2.6 years, range = 18 to 92 \n31) with normal or corrected -to-normal vision, and with no history of epileptic attacks o r 93 \nneuropsychological conditions that could interfere with  the examined study effects. The sample size 94 \nrequired to derive a reliable effect was estimated based on (Wolff et al., 2017), though our estimation 95 \nwas limited by the fact that all previous work was in a WM context. One participant did not finish the 96 \nexperiment because they were unwell, and following data inspection, two participants were removed 97 \nbecause of poor data quality due to a large number of high impedance channels, and one because of 98 \nstimulus trigger issues. Thus, EEG-based analyses were conducted based on  29 participants. For 99 \nbehavioural analyses, the first four participants were excluded because of missing button press 100 \ntriggers, which, with the further exclusion of the participant who did not complete the experiment, 101 \nresulted in an analysis of 28 participants  (participants with noisy EEG data were included in the 102 \nbehavioural analysis). 103 \n Participants were  informed about the details of the experiment  in advance —including its 104 \nduration, protocol, and methods —but were left naïve with respect to the purpose and hypotheses 105 \nassociated with the presentation of visual pings. Participants provided their written consent, and after 106 \nthe experiment, they were debriefed and given information about the central manipulation and 107 \nhypothesis upon request, and they were compensated for their time with £9 per volunteered hour. The 108 \nstudy was approved by the Ethical committee of the College of Science and Engineering  of the 109 \nUniversity of Glasgow (Application number: 300210113). 110 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 4 \n 111 \nStimulus and apparatus 112 \nThe presentation of stimuli  was controlled using PsychoPy (version 2021.2.3; Peirce et al., 2019 ) 113 \nrunning on Windows  10. Stimuli were presented on a  CRT monitor (53.3 cm; 1024 by 768 pixels) 114 \noperating at a refresh rate of  60 Hz. Participants were seated in a magnetically shielded room in a 115 \nchinrest 65 cm from the screen , or at an approximately similar distance from the screen outside the 116 \nchinrest if they experienced discomfort. Throughout the experiment, a fixation cross (with a visual 117 \nangle of 0.44°) was presented in the centre of a constantly presented grey background (RGB = 128 118 \n128 128 ; Psycho Py default ). All centrally presented stimuli overrode  the fixation dot. The visual 119 \nimpulse (i.e., ping) was a single full-contrast bullseye stimulus presented at the centre of the screen 120 \nfor 200 m illiseconds (ms; w ith a diameter of  13° and 0.31° cycles per degree ). The ping  was 121 \ngenerated using MATLAB and edited using GIMP (GNU Image Manipulation Program version 122 \n2.10.32). 123 \n In the main memory task, participants learned associations between action verbs and images, 124 \nand were later prompted with the action verb to retrieve the associated image. The action verbs were 125 \nselected based on usage frequency (largely based on Linde-Domingo et al., 2019 ) and the image 126 \nstimulus set was a combination of 192 colour images collated across various royalty free databases, 127 \nincluding the Bank of Standardized Stimuli (BOSS, Brodeur et al., 2010), and the SUN database (Xiao 128 \net al., 2010). The selected 192 images were constructed to follow a nested category structure of three 129 \nembedded hierarchical levels. At the top  level, the set consisted of 96 objects and 96 scenes, which 130 \nwere in turn composed at the middle level of 48 animate and 48 inanimate objects and 48 indoor and 131 \n48 outdoor scenes. Moving down to the bottom level, each of the middle level categories branched 132 \nout into 4 categories (e.g., for animate objects: birds, insects, mammals, and marine animals), each of 133 \nwhich contained 12 specific instances (e.g., twelve specific birds). We chose this nested hierarchy of 134 \nstimulus categories because we did not know a priori what dimension of retrieved memories would be 135 \neffectively decodable, so we included multiple levels of abstraction and chose one level based on pre-136 \ndefined criteria (See Level Selection). The objects were presented on a white square matching in size 137 \nto scene images ( i.e., the visual degrees of  all stimulus categories were  13°). Key presses were 138 \nregistered using a standard QWERTY keyboard. 139 \n 140 \nProcedure 141 \nThe main experiment consisted of 8 blocks, each with an encoding, distractor, recall, and recognition 142 \nphase (Fig. 1A). In total, the main experiment lasted between  approximately 45 and 6 5 minutes 143 \ndepending on the duration of self-paced breaks and electrode impedance maintenance. Before the 144 \nmain experiment, participants were provided with a practice run that covered each phase using 145 \nexample verbs and images that were not used in the main experiment.  A standardized set of verbal 146 \ninstructions were provided to guide participants through the practice run. If the participant reported not 147 \nunderstanding the task  or if they did not give accurate responses , the practice run  and instructions 148 \nwere repeated. Then, the main experiment commenced, throughout which EEG was acquired. At the 149 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 5 \nstart of each experimental phase, a screen was presented with a reminder of the task instructions and 150 \nrequired response keys. 151 \n 152 \n 153 \nFigure 1 . Paradigm and behavioural results . (A) Experimental paradigm. The encoding phase 154 \nconsisted of a word-image pair learning task. This was followed by a distractor task intended to wash 155 \nout working memory effects. Then, during the critical recall phase, participants were cued with words 156 \nto retrieve the paired image while visual perturbations (pings) were presented in 75% of trials. In a 157 \nfourth phase, recognition performance was tested (not displayed). (B) Average performance during 158 \nthe recognition task for trials with and without pings , collapsing across blocks for each participant . 159 \nDatapoints are individual participants. (C) Average recognition performance per participant  (i.e., 160 \ncollapsing blocks). (D) Average recognition performance per block  (i.e., collapsing participants) . (E) 161 \nAverage reaction time during encoding for subsequently recognized and forgotten trials, collapsing 162 \nacross blocks. Note: in B, C, and D, the y-axis is truncated due to high recognition performance. 163 \n 164 \n 165 \nPing No ping\n0.8\n0.9\n1\nrecognition [%]\n2468\nBlock number\nAverage\nper blockAverage\nper participant\nAverage\nper condition\n51 0 1 52 0 2 5\nParticipant\nEncoding EEG\nRecall\nWord-image pairs\n+\njump\n+\n+\njump\n+\nDistractor \ntask\nReaction time during encoding\nRecognized\nReaction time [ms]\nForgotten\nn=2194\nn=46\n01 0 0 0 2 0 0 0 3 0 0 0 4 0 0 0 5000\nE\nA\nDCB\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 6 \nIn the encoding phase, participants learned to build a mental association between action 166 \nverbs and paired images. First, a verb was presented for 1500 ms (white, OpenSans font). Then, after 167 \n1000 ms, the associated image was presented until the spacebar was pressed to indicate the 168 \nassociation was encoded (with a 6000 ms limit). Then, after a 1000 ms delay, the next verb was 169 \npresented. During each block’s encoding phase, 10 unique verb -image pairs were learned in one 170 \nshot. This resulted in 80 encoded pairs across the full experiment , with the images pseudo-randomly 171 \nselected from the full stimulus set  such as to maintain  an equal distribution of top -level stimulus 172 \ncategories (40 objects and 40 scenes) and fully random selection over nested  middle and bottom  173 \nlevels for each ping and no-ping condition. 174 \n The distractor phase that followed was included to flush out WM effects. Here, participants 175 \nperformed an odd-even task lasting 20 seconds. A number between 1 and 99 was presented in the 176 \ncentre of the screen (white, OpenSans font), and participants were instructed to press left key for odd 177 \nnumbers, and right key for even numbers. Following a left or right key press, the next number was 178 \npresented immediately. Participants’ average performance was displayed at the end of the distractor 179 \nphase, marked as the proportion of correct responses. This data was not further analysed. 180 \n Next in each block , the recall phase tested our central manipulation of a ping-based visual 181 \nperturbation. In this phase, participants recalled the learned verb-image associations of the encoding 182 \nphase. First, one of the ten encoded verbs was presented for 2000 ms , serving as the retrieval cue 183 \nthat prompted recall of the associated  image. In 75% of trials, a visual impulse was presented in 184 \neither of three time bin s: between 500 to 833.33 ms (“early ping”), 833.34 to 1116.67 ms (“middle 185 \nping”), or 1116.68 to 1500 ms (“late ping”) after the onset of the retrieval cue , with a uniform 186 \ndistribution of possible ping times within each bin. This window was chosen on the basis that previous 187 \nresearch on cued recall paradigms suggests this is the moment of maximum memory reinstatement 188 \n(Staresina & Wimber, 2019) . In 25% of trials, no visual impulse was presented in order to derive a 189 \nbaseline for statistical hypothesis testing. Participants pressed the  left key to indicate that they had 190 \nforgotten the image associated with the verb cue, or right key to indicate they remembered it. Key 191 \npresses only resulted in a new trial after 1700 ms following retrieval cue onset (i.e., 200 ms after the 192 \nlatest possible ping). With presses earlier than that , nothing happened . Participants were given a 193 \nvisual indication that key presses were available via disappearance of the retrieval cue  (at its offset; 194 \n2000 ms). During the recall phase, each of the 10 encoded verb -image pairs were tested four times , 195 \nresulting in 40 recall trials per block, and 320 trials in total , comprising 160 objects and 160 scenes . 196 \nWithin participants, each of the four conditions —early, middle, late, and no ping —were configured to 197 \npresent object and scene images equally often  (i.e., the top-level stimulus category), with the nested 198 \nmid and bottom -level categories randomized. The sequence of presented s timulus level catego ries, 199 \npinging conditions, and verb -image pairs was fully randomized within and across blocks  to mitigate 200 \norder effects. For the within block randomization, while the 40 recall trials were fully randomized, we 201 \nensured the same pair was never recalled twice in direct succession. 202 \n Finally, since the cued recall phase only included subjective memory judgments, a recognition 203 \nphase was included to obtain an objective measure of memory performance for the verb-image pairs. 204 \nDuring this two-alternative forced choice  task, o ne of the 10 encoded verbs was presented in the 205 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 7 \ncentre of the screen, with two images (visual angle of 7.8 °) presented underneath, one on the left-206 \nhand, and one on the right-hand side of the screen. Participants chose which of the two images was 207 \npaired with the central action verb using a left or right key press (with a 5000 ms time limit). The 208 \nlocation of the correctly paired image was randomized between the left and right location. The lure 209 \nimage was always another old image from the immediately preceding encoding phase. Each of the 10 210 \nencoded verb-image pairs was tested once in a random sequence. Note that we designed this study 211 \nto expend most of the available study time on the recall phase to maximize the statistical power of our 212 \nmain analysis, with the recognition phase serving chiefly as a basic check to ensure participants were 213 \nnot skipping through the experiment without memorizing verb-image pairs. 214 \n 215 \nEEG acquisition and preprocessing 216 \nThe data was recorded using a 64-channel passive EEG BrainVision system ( BrainAmp MR; Brain 217 \nProducts) with a sampling rate of 1000 Hz. For our recording software we used BrainVision Recorder 218 \n(Brain Products) . The 64 Ag/AgCl electrodes  were positioned in accordance with the extended 219 \ninternational 10-20 system. Due to a necessary change in the recording system, a different EEG cap 220 \ntype (EasyCap) was used for participants 1 to 14 (subset 1) and 15 to 33 (subset 2) . In the first  221 \nsubset, the ground electrode was located on the back of the head, below occipital electrode Oz, and 222 \ntwo EOG channels were used to monitor eye movements  (placed below and next to the eye ; VEOG 223 \nand HEOG). In the second subset, the ground electrode was on the midline frontal location AFz, and 224 \none EOG channel was used to measure eye movements (placed below the eye ; VEOG ). 225 \nFurthermore, the cap used in the second subset included channels FT9 and FT10. For event related 226 \npotential analyses, we included only electrodes common to both caps to enable a universal 227 \nvisualization of brain activity . Most electrode impedances were kept below 25  kiloΩ, and electrodes 228 \nwith outlier impedances were removed during preprocessing, with their  associated data interpolated 229 \n(see below). 230 \n Preprocessing was performed using FieldTrip (Oostenveld et al., 2011)  in MATLAB (the 231 \nMathWorks). First, the continuous EEG data was split up into two datasets: one with all trials epoched 232 \nrelative to retrieval cues, and one with  trials epoched relative to pings and no -ping (defined by 233 \nrandomly sampling ping times of the pinged trials, yielding so -called “pseudo-pings”). Put differently, 234 \nthe data was  locked once to 𝑡 = 0 defined as the  retrieval cue, and once to 𝑡 = 0 defined as the 235 \nmanipulation of interest or a baseline alternative. In both cases, the epoched trials were 4 seconds in 236 \nduration (-1 to 3 seconds relative to the event of interest). 237 \n Each dataset was filtered between 0.05 and 80 Hz  and downsampled to 250 Hz . Next, bad 238 \ntrials and channels with outlier impedance levels were manually removed via visual inspection . 239 \nSubsequently, eye movement and muscle artefacts were identified and removed  using ICA 240 \ndecomposition, and removed channels were interpolated using spline interpolation (with the  FieldTrip 241 \nfunction ft_scalpcurrentdensity). Finally, the data was re -referenced using a common average and a 242 \nLaplacian method (current source density), deriving separate data structures for cue-locked and ping-243 \nlocked analyses. 244 \n 245 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 8 \nBehavioural analysis 246 \nThe experiment was designed to result in high or even ceiling memory performance in order to obtain 247 \na maximal number of successfully remembered trials, and to optimally evaluate the central hypothesis 248 \nof a ping -induced decodability enhancement . We report objective performance for the memory test 249 \nconducted in the recognition phase, both across  pinging conditions (Fig. 1B),  participants (Fig. 1C), 250 \nand across blocks (Fig. 1D). We also report subjective judgments during the recall phase, quantifying 251 \nhow often participants report remembering versus forgetting the word -image pair. Reaction time (RT) 252 \nduring the recall phase is uninformative, because as described in the Procedure section, the response 253 \nkey was locked until 1700 ms after cue onset, at which point participants likely had already retrieved 254 \nthe associated image (Staresina & Wimber, 2019) . Indeed, participants reported actively waiting for 255 \nresponse buttons to become available. Thus, we instead analysed RT during the encoding phase as a 256 \nfunction of whether the word -image pair was subsequently recognized or not. These RT data were 257 \ncollapsed across participants and blocks  (Fig. 1E). For the proceeding analyses, both subsequently 258 \nrecognized and forgotten trials were included. 259 \n 260 \nERP Analysis 261 \nFor the ERP analyses, only channels common to both electrode cap subsets were used . We applied 262 \ntwo types of ERP analyses, one locked to (pseudo -)pings and one to retrieval cues.  FieldTrip was 263 \nused to downsample the data to 250 Hz and a band-pass filter between 0.2 and 40 Hz was used. The 264 \ndata was baseline -corrected from -200 ms to 0 ms from events of interest. For ERP traces, we 265 \ncalculated the average activity across posterior channels (C3, C4, P3, P4, O1, O2, Cz, Pz, Oz, CP1, 266 \nCP2, C1, C2, P1, P2, CP3, CP4, PO3, PO4, PO7, PO8, CPz, POz ). For ERP topographies, we used 267 \nthe 61 channels common to both ERP cap types. We statistically evaluated whether pings resulted in 268 \nhigher amplitude ERPs compared to no -ping trials using non-parametric Monte Carlo permutation 269 \ntests applied to each channel, correcting for multiple comparisons using Bonferroni correction as 270 \nimplemented in FieldTrip, averaging activity from 200 to 400 ms after pseudo-pings (alpha = 0.05; 105 271 \nrandomizations). 272 \n 273 \nMVPA analysis 274 \nFor MVPA, all EEG channels available per electrode cap type were used except EOG channels.  275 \nDepending on the analysis, w e trained and tested either a multi -class LDA using FieldTrip 276 \n(ft_timelockstatistics), or a binary -class LDA using the MVPA Light  toolbox (Treder, 2020) . We 277 \nclassified EEG data re-referenced using a Laplacian transform on the basis that it accentuates local 278 \npatterns (Kayser & Tenke, 2015) . All classifier analyses were performed on the recall phase, where 279 \nour main hypothesis could be evaluated. Unless specified otherwise, analyses were carried out on the 280 \nretrieval cue-locked dataset. We downsampled the data from 250 Hz to 50 Hz by applying a moving 281 \naverage with a window length of 140 ms, moving in steps of 20 ms . During each step, a Gaussian-282 \nweighted mean was applied in which the centre data sample of the window was multiplied by 1, and 283 \nthe tail samples by 0.15 (FWHM = ~81 ms). In a subsequent step, sample by sample, the data was z-284 \nscored across channels (i.e., setting every channel to mean = 0 and standard deviation = 1), followed 285 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 9 \nby training and testing using LDA . To evaluate decoder performance, w e applied k -fold cross 286 \nvalidation (5 folds, with 25 repetitions). For binary class decoding, w e used area under the receiver 287 \noperating characteristics curve  (AUC) as a performance metric  because it adjusts for class 288 \nimbalances (Grootswagers et al., 2017; Xie & Qiu, 2007) . For multi-class decoding, where standard 289 \nAUC is unavailable, we used accuracy and factored in level-specific differences in chance levels.  To 290 \ninfer decoding performance values under the null hypothesis , depending on the analysis,  we either 291 \nused no-ping trials or ping trials  with shuffled class labels (100 1st-level permutations, each with 3 292 \nrepetitions). All analyses were restricted to the period before button presses were made (i.e., < 2000 293 \nms). 294 \n 295 \nLevel selection 296 \nWe used a multi -class LDA on no-ping trials to determine which retrieved stimulus category (top, 297 \nmiddle, or bottom level) is most robustly detectable in the data when our main experimental 298 \nmanipulation was not applied. This level was then locked in for subsequent analyses that relate to our 299 \nkey hypothesis of ping -induced decoder enhancement. We selected the level with a high baseline 300 \nperformance to offer a conservative starting point from which we could establish whether pings are a 301 \npowerful tool to further enhance decodability. However, as we will see in the results, stimulus 302 \nselection rationales matter minimally because we found no reliable level differences in the no -ping 303 \ndecoder across levels to begin with. For statistics, we performed a Wilcoxon rank sum test comparing 304 \nthe empirical and shuffled decoding performance for each level, in the way described in the next 305 \nsection. 306 \n 307 \nMain analysis 308 \nFor the statistical analysis of the main hypothesis, we used two-level permutation testing for the ping 309 \nversus shuffle decodability comparison, and a Wilcoxon ranked sum test for the ping versus no -ping 310 \ncomparison. The former approach, which is based on van Bree et al., 2022 , implemented the 311 \nfollowing algorithm in pseudo-code—applied window-by-window: 312 \n1) For each 2 nd-level permutation  (105 times): Grab one random window -specific decodability 313 \nvalue from the 1st-level distribution of the 25 permutations of each participant and average the 314 \nresult. This yields 105 permuted averages. 315 \n2) Generate one empirical p-value by calculating the percentile of the average empirical 316 \ndecoding value within the distribution of permuted averages. 317 \nThe latter approach involved taking the Wilcoxon  signed-rank test  between the distribution of 318 \nempirical decoder results and 1st-level permutation  results across participants . We opted for a 319 \nWilcoxon test over cluster -based methods because it makes minimal assumptions about the 320 \ndistribution of decoding results (Wilcoxon, 1945; Grootswagers et al., 2017). For both approaches, we 321 \nadjusted the resulting p-values across windows for their false discover y rate (FDR). Since the p -322 \nvalues are not independent across time, we applied the approach by Benjamini & Yekutieli (2001). 323 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 10 \nFinally, for ping-locked analyses we restricted statistical analyses between 0 and 500 ms from 324 \nping onset. For analyses locked to retrieval cue, we analysed 500 to 2000 ms from cue, which is the 325 \napproximate range where memory reactivation is maximal (Staresina & Wimber, 2019). 326 \n 327 \nCondition-relative decoding peaks 328 \nIn addition to our main analysis, we  carried out a presumably more sensitive analysis to evaluate the 329 \npossibility of ping-induced decoding enhancements. We reasoned that even if visual pings do not 330 \noffer an enhancement of LTM decoding performance that is strong enough to emerge in a direct ping-331 \nto-no ping or ping -to-shuffle comparison, there could still be a weaker effect that is  detectable by 332 \nfactoring in the relative order of decoding peaks across pinging  conditions. Specifically, we tested 333 \nwhether trials with an early, middle, and late ping tended to have, respectively, earlier, later, and even 334 \nlater decoding performance peaks. In other words, we tested to what extent decoding peaks captured 335 \nping presentation order s (see Linde-Domingo et al., 2019; Mirjalili et al., 2021  for similar peak 336 \nselection approaches). 337 \nFirst, we took every participant’s SOA -specific decod ing time series —early, middle, and 338 \nlate—and extracted one peak (specified below). Then, we calculated a peak order distance (POD) per 339 \nparticipant, defined as the absolute serial distance between the order of extracted peaks and true ping 340 \npresentation order, given by the formula: 341 \n 342 \n𝑃𝑂𝐷!\"!#!\"$%&'()*+ = ' 𝑎𝑏𝑠(𝑝𝑒𝑎𝑘 − \t𝑡𝑟𝑢𝑒) 343 \n 344 \nFor example, if the decoder peak came first for early ping trials  (1 − 1), third for middle pings trials  345 \n(3 − 2), and second for late ping trials (2 − 3), this would amount to a POD of two. We divided PODs 346 \nby the maximum distance (4), normalizing the score between zero and one:  347 \n 348 \n𝑃𝑂𝐷 = ∑ 𝑎𝑏𝑠(𝑝𝑒𝑎𝑘 − \t𝑡𝑟𝑢𝑒)\n𝑚𝑎𝑥𝑖𝑚𝑢𝑚\t𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 349 \n 350 \nOn this distance metric, lower values indicate a closer correspondence between ping -induced peaks 351 \nand condition presentation order, which in turn confers stronger evidence for ping -based decoding 352 \nenhancement. For our statistical evaluation, we used a two-level permutation approach (similar to van 353 \nBree et al., 2022). Specifically, we compared the distribution of empirical PODs with PODs calculated 354 \nacross 106 second-level permutations, randomly grabbing from the pool of first-level shuffled decoder 355 \ntime courses. The p-values were defined by the resulting percentile of the empirical POD within the 356 \ndistribution of second-level shuffled PODs (one-sided test, empirical < permuted). 357 \nFor the detection of decoder peaks  in this analysis , we detected the maximum peak in the 358 \nderivative of the cumulative sum of decoding time series. We chose this peak detection method over 359 \nmore standard approaches—such as simply extracting the largest peak from raw decoding series —360 \nbecause independent simulations revealed that this algorithm is most powerful at detecting true POD 361 \neffects, outperforming a range of competing approaches (Supplementary Materials; Section 2). 362 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 11 \n 363 \nResults 364 \nBehavioural results 365 \nAs expected in light of  our experimental design, participants achieved high memory recognition 366 \nperformance, with scores approaching ceiling across behavioural analyses . First, we found no 367 \nsignificant difference in memory performance across participants between the ping (M = 0.980, SE = 368 \n0.0032) and no ping condition (M = 0.984, SE = 0.006) during the recognition phase (t(27) = -0.745, p 369 \n= 0.463; Fig. 1B ), suggesting that the decoding analyses that follow are not influenced by absolute 370 \ninter-condition differences in behaviour. This general near-ceiling performance is also apparent when 371 \nanalysing recognition performance across participants (M = 0.980, SD = 0.015; Fig. 1C) and blocks 372 \n(M = 0.980, SD = 0.009; Fig. 1D) . Furthermore, participants reported a high rate of  remembered to 373 \nforgotten judgments during the recall phase (M = 0.819; SD = 0.022). The average RT during 374 \nencoding was 2313 ms for subsequently recognized trials (SD = 1041 ms; n = 2194 trials), and 2472 375 \nms for subsequently forgotten trials (SD = 1105 ms; n = 46 trials; Fig. 1E). 376 \n 377 \nEvent-related potentials 378 \nWe observed a robust evoked EEG response after pings (Fig. 2). Specifically, for each of the three  379 \nstimulus onset asynchrony ( SOA) conditions, w e observed an extended peak of activity across 380 \noccipitoparietal channels that followed the distribution of  ping times  for retrieval cue -locked data , 381 \npeaking approximately 200 to 300 ms after pings. To further confirm that pings successfully evoked a 382 \nvisual response, we applied a ping -locked analysis across all channels and found significantly higher 383 \nERP amplitudes after pinged than no-pinged trials in posterior channels (Fig. 2, insets). Together, the 384 \nERP analysis suggests pings yielded a strong time-locked response that could putatively interact with 385 \nongoing LTM representations. For cue-locked and ping -locked ERPs for each participant, time -386 \nresolved topographical plots, and for p -values of each channel in Fig . 2 inset topographies, see the 387 \nSupplementary Materials (Section 1). 388 \n 389 \n 390 \namplitude\nEarly ping\np < 0.05\nMiddle ping\nLate ping\ncue-lock\n00 .5 11 .5 2\ntime [s]\n-1\n 0\n1\n2\n3\n4\njump\nERP per ping condition\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 12 \nFigure 2. Ping-induced e vent-related potential . Average evoked response  in posterior EEG 391 \nchannels across early (turquoise), middle (blue), and late ping (purple) trials during the recall phase. 392 \nThe inset topographies reveal higher posterior amplitudes following ping trials as contrasted with no-393 \nping trials (Monte Carlo permutation test; Bonferroni-corrected). 394 \n 395 \nDecoding results 396 \nStimulus category selection 397 \nWe used a multi -class LDA on no -ping trials (25% of the overall recall trials) to determine which 398 \nretrieved stimulus category (top, middle, or bottom level) is most robustly decodable when our main 399 \nexperimental pinging manipulation was not present (Fig. 3). We found that none of the three levels 400 \ndisplayed significant windows of decodability during our retrieval period of interest  from 500 to 2000 401 \nms after cue onset (Wilcoxon signed-rank test; p > 0.11 for top; p > 0.25 for middle; p > 0.07 for bot). 402 \nWe proceeded with the top -level, which with its two classes (objects and scenes) afforded simple 403 \nbinary classification with comparatively low variability in decoding performance. Next, during our main 404 \nanalysis, we investigated whether pings enhance the decodability of LTM contents. 405 \n 406 \n 407 \nFigure 3. Stimulus category selection. Average decoding accuracy across stimulus category levels 408 \n(top, middle, bottom). Decoding accuracy was quantified relative to the average performance across 409 \nshuffled decoding results. No significant differences were observed for any level ( Wilcoxon signed 410 \nrank test, controlled for multiple comparisons using FDR).  411 \n 412 \nMain analysis 413 \nFor our central analysis, we compared decoder performance between ping and no -ping trials for top-414 \nlevel (objects vs scenes)  classification, both with the data locked to retrieval cues, and  to 415 \npings/pseudo-pings (i.e., artificial markers derived from the pool of ping timings; Fig. 4). For the cue -416 \nlocked analysis, we found no windows where decoding was above chance for no-ping trials (two-level 417 \nMonte Carlo permutation; p > 0.49; Fig. 4A),  while the ping trials showed several significant windows 418 \nof content decodability (p < 0.05; Fig. 4B). To validate our analysis we carried out a direct comparison 419 \n-0.5 0 0.5 11 . 52\ntime [s]\n-0.05\nµshuffle\n0.05\n0.1\naccuracy\nNo ping decoder across levels\nTop\nMiddle\nBottom\ncue-lock\njump\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 13 \nbetween the ping and no-ping trial decoder, as opposed to contrasting each condition with a shuffled 420 \nbaseline. In this analysis, we found no evidence for a ping-induced decodability enhancement; neither 421 \nin the cue-locked (Wilcoxon signed-rank test; p > 0.99; Fig. 4C) nor in the (pseudo -)ping-locked data 422 \n(p > 0.99; Fig. 4D). 423 \nIn light of an important  methodological observation, we place more importance on the latter 424 \nanalysis, which directly compares the empirical decoding performance for ping and no-ping conditions 425 \nwithout leveraging shuffled results . Specifically, we observed that the standard error of the mean 426 \n(SEM) of the shuffled distributions varies substantially between ping (μSEM = 0.047) and no-ping (μSEM 427 \n= 0.028), which we speculated could be explained by trial number differences alone. We inferred that 428 \nsince the ping trial decoder was trained and tested on three times more trials than the no -ping trial 429 \ndecoder, this might naturally shrink SEM values of the shuffled distribution and thereby modulate test 430 \nstatistics. In support of this interpretation, we built a simulation which confirms that an increase in the 431 \nnumber of trials (and the number of decoding classes) reduces p-values, but only if there is an effect 432 \nin the data (Supplementary Materials; Section 3). Therefore, instead of relying on ping -to-shuffle and 433 \nno-ping-to-shuffle comparisons where power differences might misleadingly lead us to infer  a ping-434 \nrelated enhancement, we placed most credence in the  direct comparison between ping and no -ping 435 \ntrials in which shuffled results are sidestepped (Fig. 4C & Fig. 4D ; see the Supplementary Materials 436 \nfor an extended discussion; Section 3.3). 437 \n 438 \n 439 \nFigure 4. Main decoder analysis . (A) Cue-locked decoding across no-ping trials compared with a 440 \nshuffled baseline. (B) Cue-locked decoding across ping trials  compared with a shuffled baseline . (C) 441 \nDirect comparison between on ping and no-ping trials. (D) Same as (C), but with the data time-locked 442 \nPing decoderNo ping decoder\nPing > no ping decoder Ping > no ping decoder\n-0.5 0.5 11 . 52\ntime [s]\n0.46\n0.48\n0.5\n0.52\n0.54\n0.56\n0.58\ncue-lock\njump\n0\nNo ping (HA)\nShuffle (H0)\narea under curvearea under curve\nPing\nNo ping\n-0.5 0 0.5 11 . 52\ntime [s]\n0.46\n0.48\n0.5\n0.52\n0.54\n0.56\n0.58\nPing (HA)\nShuffle (H0)\np < 0.05\n-0.5 0 0.5 1\n0.48\n0.5\n0.52\n0.54\n0.56\n0.48\n0.5\n0.52\n0.54\n0.56\n-0.5 0 0.5 11 . 5 2\ncue-lock\njump\nping-lockcue-lock\njump\ntime [s] time [s]\nBA\nDC\nPing\nNo ping\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 14 \nto pings and (artificially marked) pseudo-pings. In (A) and (B) the shaded area represents the 5 th and 443 \n95th percentile of the distribution of 2nd-level permutations of the shuffled decoder, and in (B) and (C) it 444 \nrepresents the SEM of the empirical decoder . In (A) and (B), p -values were derived using two -level 445 \nMonte Carlo permutations, and in (C) and (D) using Wilcoxon signed -rank test (a ll p -values were 446 \ncorrected using FDR). 447 \n 448 \nCondition-relative decoding peaks 449 \nNext, we turn  to the presumably more sensitive  peak-order analyses. Qualitatively, we observe no  450 \nordered structure in decoder peaks when averaging across participants for each SOA pinging 451 \ncondition (Fig. 5A). For a  quantitative analysis, we formally compared peak order structure by 452 \ncomparing POD scores for the empirical and shuffled decoder using two-level permutation tests. This 453 \nanalysis confirmed the previous result by revealing no significant evidence for the hypothesis that 454 \npings induce systematic differences in the order of decoding peaks (p = 0.357; Fig. 5B). 455 \n 456 \n 457 \nFigure 5. Condition-relative peak analysis. (A) Decoding results specific to for early (cyan), middle 458 \n(blue), and late (purple) ping conditions, averaged across participants. (B) Peak order distance scores 459 \nfor the empirical decoder (red line) among a pool of 2 nd-level permutations derived from the shuffled 460 \ndecoder (grey distribution). 461 \n 462 \nDiscussion 463 \nIn this study, we set out to systematically evaluate visual perturbation, or ping-based stimulation, as a 464 \nmethod to dynamically enhance the decodability of reactivated neural representations during memory 465 \nrecall. Such an approach could supplement offline analytical approaches by adding further read -out 466 \nenhancements online at the experiment side. Despite promising results in the WM literature, in this 467 \nLTM context we found no evidence for a ping -based enhancement across several time -resolved 468 \ndecoding analyses. While pings evoked a strong brain response, they did not detectably boost neural 469 \nsignatures of memory representations  in EEG data . We draw this conclusion based on two  key 470 \nresults. First, in the main comparison between pinged trials and non -pinged trials, we found no 471 \nsignificant decoding difference regardless of whether the data was locked  to (pseudo -)pings or 472 \nretrieval cues. Second, in a more advanced analysis that leverages the constraining information of 473 \npeak order distance (POD)\n0.2 0.4 0.6 0.8 1\n2\n4\n6\n8Density (a.u.)\nRelative decoding peak order\np = 0.357\nEmpirical score\nShuffle distribution\n0.5 1 1.5 2\ntime [s]\n0.48\n0.5\n0.52\n0.54\n0.56area under curve\nEarly ping\nMiddle ping\nLate ping\nDecoder results by ping SOA\nBA\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 15 \nping presentation timings during the experiment, we also found no evidence for ping-related decoding 474 \nincreases. 475 \n There are three overarching explanations for these null results. First, there could be an effect 476 \nin the data that was left undetected analytically or statistically. Second, there could be an effect that 477 \nmanifests across other experimental contexts, but not with this study’s parameters. Third, there could 478 \nbe no effect in principle, with LTM-based retrieval eluding the enhancement of representational 479 \nreadouts using pings. We consider each option in turn. 480 \n First, the signal analysis  parameter space is high , with variability in parameters across 481 \npreprocessing and statistical analysis steps potentially altering the  results. One important source of 482 \nvariability concerns  the implementation of decoding techniques . Namely, we do not  rule out that 483 \nuntested decoding methods such as linear approaches beyond LDA  or non-linear classifiers would 484 \nhave resulted in performance enhancements induced by pings. More trivially, our analyses could have 485 \nbeen optimal, with our key statistical results containing a type-II statistical error. 486 \n Second, the parameter space  on the experimental side  is also high.  Here, we opted for a 487 \nword-image association task, which has previously been shown to afford classification -based 488 \ninferences about memory processing in the brain  (Linde-Domingo et al., 2019; Martín -Buro et al., 489 \n2020; Mirjalili et al., 2021; Kerrén et al., 2022) . However, other LTM tasks might be better suited to 490 \nreveal ping-based enhancements. Besides the memory task itself, a key set of parameters concerns 491 \nthe presentation of pings. In this study, we chose a high-intensity, short-lasting ping presented with a 492 \nuniform distribution between 500 and 1500 ms after retrieval cues. This time window was selected 493 \nbased on a review of the timeline of memory reactivation during cued recall, which suggest ed a 494 \nmaximal content reinstatement within this period (Staresina & Wimber, 2019). However, we observed 495 \nthat decoding was highest late within and even after this range, at approximately 1200 – 2000ms after 496 \ncue (see Fig. 4D).  Decoding plateaus that exceed 1500ms have also been observed in  recent work 497 \nthat employed a similar task and analysis pipeline (Kerrén et al., 2022). This raises the possibility that 498 \nthe aforementioned 500 to 15 00 ms window is  biased to be too early —perhaps because it was 499 \nestimated based on intracranial EEG research where recordings tend to focus on the hippocampus 500 \nand other regions that activate early  during retrieval (Merkow et al., 2015; Mormann et al., 2005; 501 \nStaresina et al., 2019). Put differently, it is possible that we did not find significant effects because the 502 \nsignatures of retrieved contents tended to arise robustly only after our ping presentation times . We 503 \nrecommend that future work considers later ping times, potentially informed by maximum decodability 504 \nperiods found in this and other work , or ideally in newly acquired pilot data . Moreover, additional 505 \nresearch could explore parameters such as ping duration, intensity, and strength. Furthermore, 506 \nbesides visual pings, a plethora of other perturbational approaches are on stock that could realize the 507 \nping’s proposed effects . Also inspired by WM research, stimulation using  auditory impulses might 508 \noffer a multimodal route to improving the readout  of LTM contents (Kandemir & Akyürek, 2023) . 509 \nFurthermore, brain stimulation methods like transcranial magnetic and ultrasound stimulation have the 510 \npotential to regularize brain activity through the induction of a dynamics -altering magnetic or 511 \nultrasound pulse (Moliadze et al., 2003; Mueller et al., 2014).  512 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 16 \n A third possibility is that none of these factors explain our null results, with ping-based 513 \napproaches restricting their utility to WM tasks. One specific possibility could be  that WM and LTM 514 \ndiffer in their mechanisms of action, with separate kinds of neural processes underpinning them. 515 \nIndeed, classically WM is believed to involve the active maintenance of stimulus-induced information 516 \n(Fuster & Alexander, 1971; Goldman -Rakic, 1995) , whereas LTM is assumed to be based on a  517 \ngenerative reconstruction of past experience based on the activation of silent information-storing 518 \nengrams (Josselyn & Tonegawa, 2020) . Perhaps the sweep of activity associated with the ping 519 \ninteracts more effectively with functional brain activity maintained continuously from stimulus onset , 520 \nthus explaining WM-to-LTM differences. Speaking against this interpretation is work that suggests  521 \nWM representations are encoded in activity -silent networks through  short-lasting synaptic changes 522 \n(Kamiński & Rutishauser, 2020; Masse et al., 2020; Stokes, 2015), which would not be fundamentally 523 \ndifferent from how LTM works . Contradicting this in turn is a critique  which argues that evidence for 524 \nactivity-silent networks in WM tasks  could alternatively be explained by LTM processes kicking in  525 \n(Beukers et al., 2021). Thus, since it is both unclear to what extent the mechanisms of WM and LTM 526 \ndiffer and to what extent WM and LTM intertwine in studies where ping -based effects have been 527 \ndemonstrated, we avoid firm interpretations in this part of the possibility space. In summary, although 528 \npings unambiguously elicited expected patterns of visual activity (Fig. 2), we failed to find effects on 529 \nmemory decoding, either because they were left undetected in our analysis, because they do not 530 \nshow up in our experimental protocol, or because they do not exist. 531 \n This study builds on decoding research that investigates the physical basis of memory, 532 \nleveraging it s findings for a strictly instrumental purpose: the systematic enhancement of LTM 533 \nreadouts. This undertaking is key because the field presently lacks temporally sensitive neuroimaging 534 \nmethods that enable the consistent and clear readout of memory representations, which is needed to 535 \nexplain how the brain implements memory processes.  Furthermore, the analytical challenges, null 536 \nresults, and possible solutions considered in this work could inform practice in fields closely aligned 537 \nwith memory, such as the neuroscience of mental imagery (Dijkstra et al., 2018). 538 \nTo conclude,  most efforts to improve memory readouts from electrophysiology data have  539 \nbeen restricted to  the signal analysis end. Here, we advocate for research that explores online 540 \nmanipulations as memory tasks are  unfolding, which has previously shown to complement or 541 \nsynergize with decoding techniques. For long -term memory decoding in particular however, such 542 \ninterventions are scarce, which limits research because memory involves low decodability to begin 543 \nwith. Thus, even if a further carving out of the parameter space does not demonstrate a notable 544 \nbenefit of visual perturbations, future research should creatively explore alternative online methods 545 \nsuch as multimodal stimulation and non-invasive brain stimulation. 546 \n 547 \nAcknowledgments 548 \nWe thank David Rose, Janvi Sidhu, and Jacqueline McDiarmid for their assistance during data  549 \nacquisition. This work was supported by a Starting Grant from the European Research Council 550 \nawarded to MW (ERC-2016- StG-715714). 551 \n 552 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 17 \nCompeting interests 553 \nThe authors declare no competing interests. 554 \n 555 \nReferences 556 \nBarbosa, J., Lozano -Soldevilla, D., & Compte, A. (2021). Pinging the brain with visual impulses 557 \nreveals electrically active, not activity -silent, working memories. PLoS Biology , 19(10), 558 \ne3001436. https://doi.org/10.1371/journal.pbio.3001436 559 \nBenjamini, Y., & Yekutieli, D. (2001). The control of the false discovery rate in multiple testing under 560 \ndependency. The Annals of Statistics , 29(4), 1165 –1188. 561 \nhttps://doi.org/10.1214/aos/1013699998 562 \nBeukers, A. O., Buschman, T. J., Cohen, J. D., & Norman, K. A. (2021). Is Activity Silent Working 563 \nMemory Simply Episodic Memory? Trends in Cognitive Sciences , 25(4), 284 –293. 564 \nhttps://doi.org/10.1016/j.tics.2021.01.003 565 \nBrodeur, M. B., Dionne -Dostie, E., Montreuil, T., & Lepage, M. (2010). The Bank of Standardized 566 \nStimuli (BOSS), a New Set of 480 Normative Photos of Objects to Be Used as Visual Stimuli 567 \nin Cognitive Research. PLOS ONE , 5(5), e10773. 568 \nhttps://doi.org/10.1371/journal.pone.0010773 569 \nCruzat, J., Torralba, M., Ruzzoli, M., Fernández, A., Deco, G., & Soto-Faraco, S. (2021). The phase of 570 \nTheta oscillations modulates successful memory formation at encoding. Neuropsychologia, 571 \n154, 107775. https://doi.org/10.1016/j.neuropsychologia.2021.107775 572 \nDanker, J. F., & Anderson, J. R. (2010). The ghosts of brain states past: Remembering reactivates 573 \nthe brain regions engaged during encoding. Psychological Bulletin , 136(1), 87 –102. 574 \nhttps://doi.org/10.1037/a0017937 575 \nDeuker, L., Olligs, J., Fell, J., Kranz, T. A., Mormann, F., Montag, C., Reuter, M., Elger, C. E., & 576 \nAxmacher, N. (2013). Memory Consolidation by Replay of Stimulus -Specific Neural Activity. 577 \nJournal of Neuroscience , 33(49), 19373 –19383. https://doi.org/10.1523/JNEUROSCI.0414-578 \n13.2013 579 \nDijkstra, N., Mostert, P., Lange, F. P. de, Bosch, S., & van Gerven, M. A. (2018). Differential temporal 580 \ndynamics during visual imagery and perception. eLife, 7, e33904. 581 \nhttps://doi.org/10.7554/eLife.33904 582 \nDuncan, D. H., van Moorselaar, D., & Theeuwes, J. (2023). Pinging the brain to reveal the hidden 583 \nattentional priority map using encephalography. Nature Communications , 14(1), Article 1. 584 \nhttps://doi.org/10.1038/s41467-023-40405-8 585 \nFavila, S. E., Kuhl, B. A., & Winawer, J. (2022). Perception and memory have distinct spatial tuning 586 \nproperties in human visual cortex. Nature Communications , 13(1), 5864. 587 \nhttps://doi.org/10.1038/s41467-022-33161-8 588 \nFavila, S. E., Lee, H., & Kuhl, B. A. (2020). Transforming the Concept of Memory Reactivation. 589 \nTrends in Neurosciences, 43(12), 939–950. https://doi.org/10.1016/j.tins.2020.09.006 590 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 18 \nFritch, H. A., MacEvoy, S. P., Thakral, P. P., Jeye, B. M., Ross, R. S., & Slotnick, S. D. (2020). The 591 \nanterior hippocampus is associated with spatial memory encoding. Brain Research, 1732, 592 \n146696. https://doi.org/10.1016/j.brainres.2020.146696 593 \nFuster, J. M., & Alexander, G. E. (1971). Neuron Activity Related to Short -Term Memory. Science, 594 \n173(3997), 652–654. https://doi.org/10.1126/science.173.3997.652 595 \nGoldman-Rakic, P. S. (1995). Cellular basis of working memory. Neuron, 14(3), 477 –485. 596 \nhttps://doi.org/10.1016/0896-6273(95)90304-6 597 \nGrootswagers, T., Wardle, S. G., & Carlson, T. A. (2017). Decoding Dynamic Brain Patterns from 598 \nEvoked Responses: A Tutorial on Multivariate Pattern Analysis Applied to Time Series 599 \nNeuroimaging Data. Journal of Cognitive Neuroscience , 29(4), 677 –697. 600 \nhttps://doi.org/10.1162/jocn_a_01068 601 \nHaque, R. U., Wittig, J. H., Damera, S. R., Inati, S. K., & Zaghloul, K. A. (2015). Cortical Low -602 \nFrequency Power and Progressive Phase Synchrony Precede Successful Memory Encoding. 603 \nJournal of Neuroscience , 35(40), 13577 –13586. https://doi.org/10.1523/JNEUROSCI.0687-604 \n15.2015 605 \nHaxby, J. V., Connolly, A. C., & Guntupalli, J. S. (2014). Decoding Neural Representational Spaces 606 \nUsing Multivariate Pattern Analysis. Annual Review of Neuroscience , 37(1), 435 –456. 607 \nhttps://doi.org/10.1146/annurev-neuro-062012-170325 608 \nJosselyn, S. A., & Tonegawa, S. (2020). Memory engrams: Recalling the past and imagining the 609 \nfuture. Science, 367(6473), eaaw4325. https://doi.org/10.1126/science.aaw4325 610 \nKamiński, J., & Rutishauser, U. (2020). Between persistently active and activity -silent frameworks: 611 \nNovel vistas on the cellular basis of working memory. Annals of the New York Academy of 612 \nSciences, 1464(1), 64–75. https://doi.org/10.1111/nyas.14213 613 \nKandemir, G., & Akyürek, E. G. (2023). Impulse perturbation reveals cross -modal access to sensory 614 \nworking memory through learned associations. NeuroImage, 274, 120156. 615 \nhttps://doi.org/10.1016/j.neuroimage.2023.120156 616 \nKandemir, G., Wilhelm, S. A., Axmacher, N., & Akyürek, E. G. (2023). Maintenance of colour 617 \nmemoranda in activity-quiescent working memory states: Evidence from impulse perturbation 618 \n(p. 2023.07.03.547526). bioRxiv. https://doi.org/10.1101/2023.07.03.547526 619 \nKayser, J., & Tenke, C. E. (2015). On the benefits of using surface Laplacian (Current Source 620 \nDensity) methodology in electrophysiology. International Journal of Psychophysiology : 621 \nOfficial Journal of the International Organization of Psychophysiology , 97(3), 171 –173. 622 \nhttps://doi.org/10.1016/j.ijpsycho.2015.06.001 623 \nKerrén, C., van Bree, S., Griffiths, B. J., & Wimber, M. (2022). Phase separation of competing 624 \nmemories along the human hippocampal theta rhythm. eLife, 11, e80633. 625 \nhttps://doi.org/10.7554/eLife.80633 626 \nKragel, J. E., Ezzyat, Y., Sperling, M. R., Gorniak, R., Worrell, G. A., Berry, B. M., Inman, C., Lin, J. -627 \nJ., Davis, K. A., Das, S. R., Stein, J. M., Jobst, B. C., Zaghloul, K. A., Sheth, S. A., Rizzuto, D. 628 \nS., & Kahana, M. J. (2017). Similar patterns of neural activity predict memory function during 629 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 19 \nencoding and retrieval. NeuroImage, 155, 60 –71. 630 \nhttps://doi.org/10.1016/j.neuroimage.2017.03.042 631 \nKuhl, B. A., Rissman, J., & Wagner, A. D. (2012). Multi-voxel patterns of visual category 632 \nrepresentation during episodic encoding are predictive of subsequent memory. 633 \nNeuropsychologia, 50(4), 458–469. https://doi.org/10.1016/j.neuropsychologia.2011.09.002 634 \nLinde-Domingo, J., Treder, M. S., Kerrén, C., & Wimber, M. (2019). Evidence that neural information 635 \nflow is reversed between object perception and object reconstruction from memory. Nature 636 \nCommunications, 10(1), 179. https://doi.org/10.1038/s41467-018-08080-2 637 \nMadore, K. P., & Wagner, A. D. (2022). Readiness to remember: Predicting variability in episodic 638 \nmemory. Trends in Cognitive Sciences , 26(8), 707 –723. 639 \nhttps://doi.org/10.1016/j.tics.2022.05.006 640 \nMaguire, E. A. (2014). Memory consolidation in humans: New evidence and opportunities. 641 \nExperimental Physiology, 99(3), 471–486. https://doi.org/10.1113/expphysiol.2013.072157 642 \nMartín-Buro, M. C., Wimber, M., Henson, R. N., & Staresina, B. P. (2020). Alpha Rhythms Reveal 643 \nWhen and Where Item and Associative Memories Are Retrieved. The Journal of 644 \nNeuroscience: The Official Journal of the Society for Neuroscience , 40(12), 2510 –2518. 645 \nhttps://doi.org/10.1523/JNEUROSCI.1982-19.2020 646 \nMasse, N. Y., Rosen, M. C., & Freedman, D. J. (2020). Reevaluating the Role of Persistent Neural 647 \nActivity in Short -Term Memory. Trends in Cognitive Sciences , 24(3), 242 –258. 648 \nhttps://doi.org/10.1016/j.tics.2019.12.014 649 \nMerkow, M. B., Burke, J. F., & Kahana, M. J. (2015). The human hippocampus contributes to both the 650 \nrecollection and familiarity components of recognition memory. Proceedings of the National 651 \nAcademy of Sciences, 112(46), 14378–14383. https://doi.org/10.1073/pnas.1513145112 652 \nMirjalili, S., Powell, P., Strunk, J., James, T., & Duarte, A. (2021). Context Memory Encoding and 653 \nRetrieval Temporal Dynamics are Modulated by Attention across the Adult Lifespan. eNeuro, 654 \n8(1), ENEURO.0387-20.2020. https://doi.org/10.1523/ENEURO.0387-20.2020 655 \nMoliadze, V., Zhao, Y., Eysel, U., & Funke, K. (2003). Effect of transcranial magnetic stimulation on 656 \nsingle-unit activity in the cat primary visual cortex. The Journal of Physiology, 553(Pt 2), 665–657 \n679. https://doi.org/10.1113/jphysiol.2003.050153 658 \nMormann, F., Fell, J., Axmacher, N., Weber, B., Lehnertz, K., Elger, C. E., & Fernández, G. (2005). 659 \nPhase/amplitude reset and theta –gamma interaction in the human medial temporal lobe 660 \nduring a continuous word recognition memory task. Hippocampus, 15(7), 890 –900. 661 \nhttps://doi.org/10.1002/hipo.20117 662 \nMueller, J., Legon, W., Opitz, A., Sato, T. F., & Tyler, W. J. (2014). Transcranial Focused Ultrasound 663 \nModulates Intrinsic and Evoked EEG Dynamics. Brain Stimulation , 7(6), 900 –908. 664 \nhttps://doi.org/10.1016/j.brs.2014.08.008 665 \nOostenveld, R., Fries, P., Maris, E., & Schoffelen, J. -M. (2011). FieldTrip: Open source software for 666 \nadvanced analysis of MEG, EEG, and invasive electrophysiological data. Computational 667 \nIntelligence and Neuroscience, 2011, 156869. https://doi.org/10.1155/2011/156869 668 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 20 \nPearson, J., Naselaris, T., Holmes, E. A., & Kosslyn, S. M. (2015). Mental Imagery: Functional 669 \nMechanisms and Clinical Applications. Trends in Cognitive Sciences , 19(10), 590 –602. 670 \nhttps://doi.org/10.1016/j.tics.2015.08.003 671 \nPeirce, J., Gray, J. R., Simpson, S., MacAskill, M., Höchenberger, R., Sogo, H., Kastman, E., & 672 \nLindeløv, J. K. (2019). PsychoPy2: Experiments in behavior made easy. Behavior Research 673 \nMethods, 51(1), 195–203. https://doi.org/10.3758/s13428-018-01193-y 674 \nRissman, J., & Wagner, A. D. (2012). Distributed representations in memory: Insights from functional 675 \nbrain imaging. Annual Review of Psychology , 63, 101–128. https://doi.org/10.1146/annurev-676 \npsych-120710-100344 677 \nRizzuto, D. S., Madsen, J. R., Bromfield, E. B., Schulze -Bonhage, A., Seelig, D., Aschenbrenner -678 \nScheibe, R., & Kahana, M. J. (2003). Reset of human neocortical oscillations during a working 679 \nmemory  task. Proceedings of the National Academy of Sciences , 100(13), 7931 –7936. 680 \nhttps://doi.org/10.1073/pnas.0732061100 681 \nSchreiner, T., Petzka, M., Staudigl, T., & Staresina, B. P. (2021). Endogenous memory reactivation 682 \nduring sleep in humans is clocked by slow oscillation -spindle complexes. Nature 683 \nCommunications, 12(1), Article 1. https://doi.org/10.1038/s41467-021-23520-2 684 \nStaresina, B. P., Reber, T. P., Niediek, J., Boström, J., Elger, C. E., & Mormann, F. (2019). 685 \nRecollection in the human hippocampal -entorhinal cell circuitry. Nature Communications , 686 \n10(1), 1503. https://doi.org/10.1038/s41467-019-09558-3 687 \nStaresina, B. P., & Wimber, M. (2019). A Neural Chronometry of Memory Recall. Trends in Cognitive 688 \nSciences, 23(12), 1071–1085. https://doi.org/10.1016/j.tics.2019.09.011 689 \nStokes, M. G. (2015). “Activity -silent” working memory in prefrontal cortex: A dynamic coding 690 \nframework. Trends in Cognitive Sciences , 19(7), 394 –405. 691 \nhttps://doi.org/10.1016/j.tics.2015.05.004 692 \nTen Oever, S., De Weerd, P., & Sack, A. T. (2020). Phase -dependent amplification of working 693 \nmemory content and performance. Nature Communications , 11(1), 1832. 694 \nhttps://doi.org/10.1038/s41467-020-15629-7 695 \nTer Wal, M., Linde-Domingo, J., Lifanov, J., Roux, F., Kolibius, L. D., Gollwitzer, S., Lang, J., Hamer, 696 \nH., Rollings, D., Sawlani, V., Chelvarajah, R., Staresina, B., Hanslmayr, S., & Wimber, M. 697 \n(2021). Theta rhythmicity governs human behavior and hippocampal signals during memory -698 \ndependent tasks. Nature Communications, 12(1), 7048. https://doi.org/10.1038/s41467-021-699 \n27323-3 700 \nTreder, M. S. (2020). MVPA -Light: A Classification and Regression Toolbox for Multi -Dimensional 701 \nData. In Frontiers in Neuroscience  (Vol. 14). 702 \nhttps://www.frontiersin.org/article/10.3389/fnins.2020.00289 703 \nvan Bree, S., Melcón, M., Kolibius, L. D., Kerrén, C., Wimber, M., & Hanslmayr, S. (2022). The brain 704 \ntime toolbox, a software library to retune electrophysiology data to brain dynamics. Nature 705 \nHuman Behaviour, 6(10), 1430–1439. https://doi.org/10.1038/s41562-022-01386-8 706 \nWilcoxon, F. (1945). Individual Comparisons by Ranking Methods. Biometrics Bulletin, 1(6), 80–83. 707 \nhttps://doi.org/10.2307/3001968 708 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 21 \nWolff, M. J., Ding, J., Myers, N. E., & Stokes, M. G. (2015). Revealing hidden states in visual working 709 \nmemory using electroencephalography. Frontiers in Systems Neuroscience , 9. 710 \nhttps://doi.org/10.3389/fnsys.2015.00123 711 \nWolff, M. J., Jochim, J., Akyürek, E. G., Buschman, T. J., & Stokes, M. G. (2020). Drifting codes within 712 \na stable coding scheme for working memory. PLOS Biology , 18(3), e3000625. 713 \nhttps://doi.org/10.1371/journal.pbio.3000625 714 \nWolff, M. J., Jochim, J., Akyürek, E. G., & Stokes, M. G. (2017). Dynamic hidden states underlying 715 \nworking-memory-guided behavior. Nature Neuroscience , 20(6), 864 –871. 716 \nhttps://doi.org/10.1038/nn.4546 717 \nXiao, J., Hays, J., Ehinger, K. A., Oliva, A., & Torralba, A. (2010). SUN database: Large -scale scene 718 \nrecognition from abbey to zoo. 2010 IEEE Computer Society Conference on Computer Vision 719 \nand Pattern Recognition, 3485–3492. https://doi.org/10.1109/CVPR.2010.5539970 720 \nXie, J., & Qiu, Z. (2007). The effect of imbalanced data sets on LDA: A theoretical and empirical 721 \nanalysis. Pattern Recognition, 40(2), 557–562. https://doi.org/10.1016/j.patcog.2006.01.009 722 \nXue, G. (2018). The Neural Representations Underlying Human Episodic Memory. Trends in 723 \nCognitive Sciences, 22(6), 544–561. https://doi.org/10.1016/j.tics.2018.03.004 724 \nYang, C., He, X., & Cai, Y. (2023). Reactivating and reorganizing activity-silent working memory: Two 725 \ndistinct mechanisms underlying pinging the brain  (p. 2023.07.16.549254). bioRxiv. 726 \nhttps://doi.org/10.1101/2023.07.16.549254 727 \n 728 \n  729 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 22 \nSupplementary Materials 730 \n1. Event-related potential 731 \n 732 \n 733 \nSupplementary Figure 1. Retrieval cue-locked ERP . The purple trace reflects the average cue -734 \nlocked response for each participant across posterior EEG channels. The grey horizontal line 735 \nrepresents cue onset. For more details, see the Methods section in the main text.  The amplitude on 736 \nthe y-axis is in arbitrary units. 737 \n 738 \n 739 \n 740 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 23 \nSupplementary Figure 2. Retrieval ping-locked ERP. The purple trace reflects the average ping-741 \nlocked response for each participant across posterior EEG channels. The grey horizontal line 742 \nrepresents ping onset. For more details, see the Methods section in the main text.  The amplitude on 743 \nthe y-axis is in arbitrary units. 744 \n 745 \n 746 \n 747 \n 748 \nSupplementary Figure 3. Retrieval cue-locked topographies. These topographical plots  represent 749 \nthe average cue -locked activity across participants.  The colours represent the difference in EEG 750 \nactivity before and after cue onset in arbitrary units (red colours represent activity post > activitypre and 751 \nvice versa for blue colours).  No statistical analysis was carried out for these topographical contrasts. 752 \nFor more details, see the Methods section in the main text. 753 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 24 \n 754 \n 755 \nSupplementary Figure 4. Retrieval ping -locked topographies (ping vs. no ping trials) . These 756 \ntopographical plots represent the average cue -locked activity across participants. The colours 757 \nrepresent the difference in EEG activity between ping and no-ping (red colours represent activityping > 758 \nactivityno ping  and vice versa for blue colours).  No statistical analysis was carried out for these 759 \ntopographical contrasts. For more details, see the Methods section in the main text. 760 \n 761 \n 762 \nChannel Early ping (p-val) Middle ping (p-val) Late ping (p-val) \nFp1 0.032 0.616 0.246 \nFpz 0.089 0.079 0.011 \nFp2 0.042 0.537 0.115 \nAF8 0.119 0.422 0.318 \nAF7 0.014 0.272 0.954 \nAF3 0.439 0.23 0.123 \nAF4 0.712 0.541 0.014 \nF7 0.002 0.346 0.358 \nF5 0.068 0.439 0.33 \nF3 0.119 0.477 0.693 \nF1 0.597 0.662 0.119 \nFz 0.053 0.551 0.003 \nF2 0.013 0.473 0 \nF4 0.341 0.939 0.049 \nF6 0.427 0.559 0.707 \nF8 0.131 0.826 0.825 \nFT8 0.001 0.097 0.049 \nFC6 0.177 0.142 0.78 \nFC4 0.962 0.176 0.881 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 25 \nFC2 0.245 0.276 0.09 \nFC1 0.969 0.503 0.881 \nFC3 0.176 0.279 0.28 \nFC5 0.011 0.083 0.112 \nFT7 0.002 0.298 0.127 \nT7 0.004 0.047 0.043 \nC5 0.013 0.027 0.051 \nC3 0.002 0.022 0.01 \nC1 0.148 0.049 0.04 \nCz 0.144 0.155 0.114 \nC2 0.305 0.182 0.104 \nC4 0.02 0.004 0.014 \nC6 0.003 0 0.003 \nT8 0.002 0.002 0.003 \nTP10 0.002 0 0 \nTP8 0 0 0 \nCP6 0 0 0 \nCP4 0.001 0 0.001 \nCP2 0.006 0.001 0.001 \nCPz 0.019 0.006 0.01 \nCP1 0.272 0.002 0.003 \nCP3 0 0 0.001 \nCP5 0.002 0.001 0.001 \nTP7 0.002 0.008 0.001 \nTP9 0 0 0 \nP7 0 0 0 \nP5 0 0 0 \nP3 0 0 0 \nP1 0 0 0 \nPz 0.001 0 0 \nP2 0.001 0 0 \nP4 0 0 0 \nP6 0 0 0 \nP8 0 0 0 \nPO8 0 0 0 \nPO4 0 0 0 \nPOz 0 0 0 \nPO3 0 0 0 \nPO7 0 0 0 \nO1 0 0 0 \nOz 0.001 0 0 \nO2 0 0 0 \n 763 \nSupplementary Table 1. P-values associated with inset topographies in main text Fig. 2; rounded to 764 \nthree decimal points. 765 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 26 \n2. Peak order analysis simulation 766 \n2.1 Time series simulation 767 \nWe used MATLAB (the MathWorks) to generate time series with two components: (1) a peak at a 768 \nfixed time point (1000 ms), and (2) autocorrelated noise generated using a random walk  procedure. 769 \nWe matched several characteristics of the simulated time series  to our empirical decoding data, 770 \nincluding the analysis period (500 to 2000 ms) , sampling rate (50 Hz) , and the number of (virtual) 771 \nparticipants (N = 29) . The signal -to-noise (SNR) ratio of the simulation was set to 1.15 , qualitatively 772 \nmatching peaks observed in the empirical data. We found that varying the SNR does not significantly 773 \nalter the results. We generated 1000 trials per participant, resulting in 29000 trials in total. 774 \n 775 \n2.2 Analysis 776 \nWe included a smoothing parameter that implemented one of four smoothing methods : no filter, a 777 \nGaussian filter, a Savitzky-Golay filter, and a median filter. We also included a window size for 778 \nsmoothing, set to 10 samples for our main analysis. We compared the performance of eight peak 779 \ndetection methods, evaluating each of them based on the absolute distance between estimated peaks 780 \nand true peaks—amounting to a simplified version of the peak order distance score described under 781 \ncondition-relative decoding peaks in the main text . The winning method was locked in for our 782 \nempirical analysis. We tested eight peak detection methods: 783 \n(1) Low-pass approach, where the maximum peak was computed after a low -pass filter was 784 \napplied to the time series. 785 \n(2) Maximum value approach, which simply computed the maximum value per time series 786 \nregardless of whether the surrounding data was peak-like. 787 \n(3) Cumulative sum approach, which computed the maximum peak in the derivative of the 788 \ncumulative sum of the data. 789 \n(4) Cumulative integral approach, which computed the maximum peak in the cumulative integral 790 \nof the data via the trapezoidal method. 791 \n(5) Integral cumulative sum approach, which worked as the previous method but which operates 792 \nover the cumulative sum rather than raw time series. 793 \n(6) Wavelet transform-based method, which finds the maximum peak  in a wavelet decomposed 794 \nversion of the data. 795 \n(7) Hilbert transform-based method, which find the maximum peak in the amplitude fluctuations in 796 \nthe envelope of the time series. 797 \n(8) Cross-correlation method, which finds the time lag with a maximal correlation between the 798 \nsignal and iteratively shifted versions of itself. 799 \n 800 \n2.3 Results 801 \nWe found that approach 5—the i ntegral cumulative sum approach —reliably achieves low absolute 802 \ndistance errors across parameters (Supplementary Figure 5). These results were generally 803 \nunchanged across adjustments of the  parameters (to evaluate this, we refer to the code published 804 \nwith this manuscript). Thus, we used approach 5 in our main peak order detection analysis. 805 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 27 \n 806 \nSupplementary Figure 6. In simulated time series, the integral cumulative sum approach works best 807 \nfor detecting a peak in noisy time series. The red circle indicates the best -performing method, and 808 \nyellow the second best-performing method. Errors were computed based on the absolute distance in 809 \nmilliseconds (ms) between estimated and true peak location. 810 \n 811 \n3. Class and trial number decoding simulation 812 \nWe speculated based on a qualitative inspection of the empirical  decoding results that the number of 813 \ntrials (Ntrials) and classes (Nclasses) reduces the statistical significance of decoding results. We  814 \nevaluated this intuition by demonstrating using simulations that these two parameters  do indeed  815 \ninfluence the variance of shuffled and empirical results, which in turn affects p-values but only if there 816 \nis a true effect in the data. 817 \n 818 \n3.1 Time series simulation 819 \nUsing MATLAB, w e generated one ground truth vector of class labels  which represented the true 820 \nclass structure in the simulated data. This vector contained a random sequence of integers randomly 821 \ngrabbed between the interval 1 and Nclasses. For example, with 16 classes, the ground truth pattern 822 \nmight have contained a sequence of [2,7,15,4,13,17] and with 2 classes a sequence of [2,2,1,2,1,2]. 823 \n Then, to simulate shuffled decoding results, we generated a distribution of random sequences 824 \nof integers identical to the ground truth procedure, but with newly generated random integers. These 825 \nrandom sequences represented shuffled decoding results and were scored based on their average 826 \nelement-wise correspondence to the ground truth pattern —which is how decoding accuracy is  827 \nnormally computed. For example, if the permuted vector is [2,1,2,2,1,1] and the true sequence is 828 \n[2,2,1,2,1,2], the accuracy would be  50% because half of the class labels correspond to the true 829 \nstructure. Trivially, with increasing repetitions the shuffled distribution  will approach  chance level 830 \npredictions of the ground truth pattern (i.e., the expected value is exactly at 1/Nclasses). 831 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 28 \n Finally, to simulate empirical decoding results, we again generated a distribution of random 832 \nintegers identical to the procedure for shuffled and ground truth decoding results. However, for these 833 \ndata we manually injected between 0% and 60% of the ground truth  pattern into the otherwise 834 \nrandom vector, effectively modulating decoding accuracy. With 0% of the ground truth injected, there 835 \nis no statistically detectable difference in accuracy between empirical and shuffled decoding results, 836 \nbecause the vectors  are equally random. With 60%, the encoding results are substantially more 837 \naccurate than shuffled results, yielding above chance decoding accuracy. 838 \nWe simplified our simulation by operationalizing the variable Ntrials as the number of elements 839 \nin the vector , allowing us to efficiently investigate  how the number of observations influences 840 \nstatistical tests. We also compared Nclasses = 2 and Nclasses = 16, which respectively match the number 841 \nof classes for top- and bottom-level category decoding in our main experiment. Both Ntrials and Nclasses 842 \nwere independently manipulated in a 2 ∗ 2 factorial design, allowing us to evaluate the contribution of 843 \neach variable toward statistical outcomes (as a function of effect size). 844 \n 845 \n3.2 Results 846 \nFirst, with respect to N classes, we found that increasing the number of classes reduces the spread of 847 \nboth shuffled and empirical decoding results (Supplementary Figure 7; columns). This happens both if 848 \nthere is no true effect in the empirical data, and when a significant proportion of the ground truth is 849 \ninserted into the empirical data . Second, we found  that Ntrials similarly reduces the variance of both 850 \nshuffled and decoding results, both across low and high N classes (Supplementary Figure 7; top and 851 \nbottom half). Thus, we conclude that both factors modulate the likelihood of finding a significant 852 \ndifference between empirical and shuffled results, but only if there is a true effect in the data. Indeed, 853 \nas we can glean from the results based on non -existent effects, the distributions of empirical and 854 \nshuffled will overlap regardless of N trials or Nclasses (Supplementary Figure 7; left half). In contrast, if 855 \nthere is an effect (60% injected ground truth), both N trials and N classes independently increase the 856 \ndistributional distance between empirical and shuffled accuracy values.  857 \n 858 \n3.3 Discussion 859 \nWe found that N trials and N classes independently reduce the variance of accuracy results, which will 860 \naffect statistical tests between empirical and shuffled distributions but only if there is an effect in the 861 \ndata. As suggested in the main text, these findings suggest that statistical analyses that depend on 862 \nvariance comparisons between empirical and shuffled distributions should be interpreted with care if it 863 \nis done across conditions with varying Ntrials and Nclasses. With regard to our main analysis for example, 864 \nthe fact that the decoder based on pinged trials yields more significant decodability compared to the 865 \ndecoder based on no-pinged trials should be interpreted with caution because there are differences in 866 \nNtrials between the two conditions that could partially or fully explain this effect. More generally, we 867 \nfound that the condition with more trials or more classes is by default more likely to yield significant p-868 \nvalues—but only if a true effect exist. 869 \n 870 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 29 \n 871 \nSupplementary Figure 7. The effects of class and trial number on decoding accuracy. Both the 872 \nnumber of classes (columns) and trials (top vs. bottom half) influences the distance between shuffled 873 \nand empirical distributions—but only if there is an effect in the data (left vs. right half). 874 \n 875 \nThese findings may be a manifestation of the classical notion of statistical power in statistical 876 \nanalysis but within the less intuitive context of decoding accuracy . Our interpretation then is not that 877 \nNtrials and N classes must necessarily be equal between conditions for a statistical comparison to be 878 \nmeaningful. Rather, we wanted to err on the side of caution and ensure that  analyses where power 879 \ndifferences could possibly explain condition differences (e.g., Fig. 3 and Fig. 4A and 4B  in the main 880 \ntext) do not inform subsequent analyses and scientific interpretations  by themselves . Instead, we 881 \nsupplemented each of the implicated analyses with additional rationale (in the case of Fig. 3) or 882 \nanalyses that do not involve empirical -to-shuffle decoding comparisons. Indeed, Fig. 4C and Fig. 4D 883 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint \n\n 30 \ninvolve direct comparisons between empirical and shuffled distributions , sidestepping the issue  884 \naltogether. 885 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted April 23, 2024. ; https://doi.org/10.1101/2024.04.19.590215doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}